apecs
Fast Entity-Component-System library for game programming
https://github.com/jonascarpay/apecs#readme
Version on this page: | 0.9.5@rev:1 |
LTS Haskell 22.36: | 0.9.6 |
Stackage Nightly 2024-10-05: | 0.9.6 |
Latest on Hackage: | 0.9.6 |
apecs-0.9.5@sha256:e077d969a6e182a754cd63d12196d0c5d2cf2ab9ff2baad8c73b8e1c7de5f2ee,2186
Module documentation for 0.9.5
apecs
apecs is an Entity Component System (ECS) library for game development.
apecs aims to be
- Fast - Performance is competitive with Rust ECS libraries (see benchmark results below)
- Safe - Completely hides the dangers of the low-level machinery
- Concise - Game logic is expressed using a small number of powerful combinators
- Flexible - Easily add new modules or backends
- Cool
Links
- documentation on hackage
- tutorial and other examples
Games/articles
- Notakto, and associated blog post/apecs tutorial by @Ashe
- mallRL - a grocery shopping roguelike by @nmaehlmann
- An implementation of the Unity tutorial project using apecs by @mewhhaha
- SpaceMar by @dpwiz
- Achtung die haskell by @mewhhaha
- (Your game here)
Packages
-
apecs-physics - 2D physics using the Chipmunk2D engine
-
apecs-gloss - Simple frontend for gloss-based rendering
-
apecs-stm - STM-based stores for easy concurrency
Example
{-# LANGUAGE FlexibleInstances #-}
{-# LANGUAGE MultiParamTypeClasses #-}
{-# LANGUAGE ScopedTypeVariables #-}
{-# LANGUAGE TemplateHaskell #-}
{-# LANGUAGE TypeFamilies #-}
import Apecs
import Linear (V2 (..))
newtype Position = Position (V2 Double) deriving Show
newtype Velocity = Velocity (V2 Double) deriving Show
data Flying = Flying
makeWorldAndComponents "World" [''Position, ''Velocity, ''Flying]
game :: System World ()
game = do
newEntity (Position 0, Velocity 1)
newEntity (Position 2, Velocity 1)
newEntity (Position 1, Velocity 2, Flying)
-- 1. Add velocity to position
-- 2. Apply gravity to non-flying entities
-- 3. Print a list of entities and their positions
cmap $ \(Position p, Velocity v) -> Position (v+p)
cmap $ \(Velocity v, _ :: Not Flying) -> Velocity (v - V2 0 1)
cmapM_ $ \(Position p, Entity e) -> liftIO . print $ (e, p)
main :: IO ()
main = initWorld >>= runSystem game
Changes
[0.9.5]
Added
- (#99)
collect
Changed
- (#84) Updated links to articles
[0.9.4]
Changed
- (#86) Add support for GHC 9.2 and Template Haskell 2.18
[0.9.3]
Added
newEntity_ = void . newEntity
[0.9.2]
Changed
- (#68) Add instances of MonadThrow, MonadCatch, and MonadMask to SystemT
- Cleaned up the README
- Small haddock fixes
[0.9.1]
Changed
- (#63) Fixed bug where
modify
on non-existent components crashes
[0.9.0]
Added
- (#59) Expose
makeMapComponents
, which createsComponent
instances withMap
stores
Changed
- (#60) Add
Component
type names in non-existent component errors - Relaxed the type of
modify
to allow a different return type - Constrain the
cmapM_
type(c -> SystemT w m ()) -> SystemT w m ()
, to make it clearer that the inner function does not updateComponents
- Simplify nix infrastructure
[0.8.3]
Changed
- (#58) Added support for Template Haskell 2.15.0.0 through CPP flags
[0.8.2]
Changed
- (#55) Fixed a bug where components where not properly deleted from the cache following the cache bitmasking update
[0.8.1]
Changed
- Changed
Cache
s to use bitmasks instead of the remainder operation. This makes caches up to three times faster. - Fixed bug in
Cache
where the same entity appeared in member list of both the cache and the underlying store
[0.8.0]
There are a number of unsolved problems in apecs’ design space. Most notably, it needs a good way to do streaming and reactivity, or find a way to integrate with existing solutions. I’m hesitant to accept some of the feature requests I’ve gotten because they would be obsoleted when we figure this out, and I don’t want to pollute the API with unstable features.
However, I don’t think we should let perfect get in the way of good.
So, apecs 0.8 have new Apecs.Experimental.*
modules, that I want to use for features that might or might not get removed or changed.
They should provide some conveniences, but the catch is that their API might undergo significant changes between point releases (and therefore within LTS’es).
Some of the existing modules were already marked experimental, and those have been moved.
Added
- Experimental
Head
component OrdMap
andIxMap
reactive maps
Changed
- Moved
ReadOnly
toApecs.Stores
sinceEntityCounter
now depends on it - Moved spatial hashing to
Experimental.Util
- Moved
Redirect
andStores.Extra
toExperimental.Stores
- Moved
Reactive
toExperimental.Reactive
rget
has been replaced bywithReactive
- Improved error messages for unsafe operations
[0.7.3]
Changed
- Added Data.Semigroup to Stores.Extra to build with GHC 8.2.2 in hackage matrix
[0.7.2]
Changed
- Fixed bug in the
Pushdown
store Apecs
module no longer re-exports the entireData.Proxy
module, but instead justProxy (..)
.- Added (approximate?) lower and upper version bounds to dependencies
[0.7.1]
Added
$=
and$~
operators as synonyms forset
andget
respectively
Removed
getAll
andcount
, which were made redundant bycfold
.
[0.7.0]
Added
- The
Reactive
store and module is a redesign of theRegister
store, and provides a more general solution for ‘stores that perform additional actions when written to’. - The
Apecs.Stores.Extra
submodule, which contains thePushdown
andReadOnly
stores.Pushdown
adds pushdown semantics to stores, andReadOnly
hides theExplSet
instances of whatever it wraps. - The
EntityCounter
and associated functions have all been specified toIO
, sinceGlobal EntityCounter
only works in IO. Furthermore,EntityCounter
now uses aReadOnly
store, to prevent users from accidentally changing its value. Redirect
component that writes to another entity incmap
.
Changed
- Default stores have
MonadIO m => m
instances, rather thanIO
. This makes it easier to nestSystemT
. - All apecs packages have been consolidated into a single git repo.
Apecs.Components
contains the components (and corresponding stores) fromApecs.Core
.
[0.6.0.0]
Changed
- Nothing, but since 0.5.1 was API-breaking I’ve decided to bump to 0.6
[0.5.1.1]
Changed
Register
needs UndecidableInstances in GHC 8.6.2, I’m looking for a way around this. I’ve removed it for now.
[0.5.1.0]
Added
- The
Register
store, which allows reverse lookups for bounded enums. For example, ifBool
has storageRegister (Map Bool)
,regLookup True
will yield a list of all entities with aTrue
component. Can also be used to emulate a hash table, wherefromEnum
is the hashing function. This allows us to make simple spatial hashes. I’m open to suggestions for better names than Register. cmapIf
, cmap with a conditional test
Changed
ExplInit
now too takes a monad argument.- Started rewrite of the test suite
- Caches now internally use -2 to denote absence, to avoid possible conflict with -1 as a global entity
Removed
- The STM instances have been removed, to be moved to their own package
[0.5.0.0]
Changed
System w a
is now a synonym forSystemT w IO a
. A variable monad argument allows apecs to be run in monads like ST or STM. Most of the library has been rewritten to be as permissive as possible in its monad argument.
Added
- STM stores. These will be moved to a separate package soon.
[0.4.1.2]
Changed
- Either can now be deleted, deleting
Either a b
is the same as deleting(a,b)
. - Some were missing their inline pragma’s, now they don’t
[0.4.1.1]
Changed
- Export
Get
,Set
,Destroy
,Members
by default - Export
cfold
,cfoldM
,cfoldM_
by default - Fix () instance
[0.4.1.0]
Added
cfold
,cfoldM
,cfoldM_
Either
instances andEitherStore
Changed
- Changed MaybeStore implementation to no longer use -1 for missing entities.
- Fixed some outdated documentation.
- Change the
global
void entity to -2, just to be sure it won’t conflict if accidentally used in a cache.
[0.4.0.0]
Added
- A changelog
Changed
Store
is now split into 5 separate type classes;ExplInit
,ExplGet
,ExplSet
,ExplDestroy
, andExplMembers
. This makes it illegal to e.g. iterate over aNot
.- phantom arguments are now given as
Proxy
values, re-exported fromData.Proxy
. This makes phantom arguments explicit and avoids undefined values.